Hosting a dashboad on my own website

matsuo_basho 10

I would like to build a dashboard using Plotly Dash and then host it on my own website. The dashboard will allow users to load data, then run machine learning models on that data in the backend and output the results.

I know Azure ML and have worked with it quite a bit, but I don't have familiarity with any of the other services in the Azure ecosystem. I searched around and seems that it's possible to do this with Web Apps or App Service, but not sure.

Please provide some guidance and links on how to get started. Also, the way I would integrate it with AzureML for the modeling portion would be helpful, as well.

Thanks.

Gowtham CP 3,730 Reputation points

2024-06-02T05:06:03.7733333+00:00

Hello matsuo_basho ,

We haven't heard back from you. Could you please provide an update on your issue? If you have any questions or concerns, feel free to reach out to us. If the information provided has been helpful, please consider marking it as the accepted answer to close this case. Doing so will assist others with similar questions in finding solutions more easily.
matsuo_basho 10 Reputation points

2024-06-02T15:57:53.3466667+00:00

@Gowtham CP thanks for checking. Can you recommend a resource that describes how to use Azure Functions for my use case. I found the following tutorial:
https://learn.microsoft.com/en-us/azure/azure-functions/functions-machine-learning-tensorflow?tabs=bash

But not sure whether it does the backend processing that I need.

Also, I'm not sure I actually need to use AzureML. The dashboard process is fairly simple - the user loads data, the Python code trains an ML model, that model for prediction, and then the results are returned to the dashboard. I'm not storing the models across sessions at this point.
Gowtham CP 3,730 Reputation points

2024-06-03T07:08:40.2233333+00:00

Hello again matsuo_basho ,

If you're building a basic dashboard with data upload and on-demand model training, skip Azure Functions and ML for now.

Just use dash-uploader for file upload in your Dash app. Train a new model (your preferred library) and make predictions on the uploaded data in the Python backend. Deploy using a web server or hosting service. This keeps it simple and gives you control over the training process!

For future scalability or complex workflows, consider using Azure Functions for backend processing. The provided link is a good starting point. This option offers more flexibility for growth.

Hope it helps! Thanks!!
matsuo_basho 10 Reputation points

2024-06-03T16:53:08.14+00:00

@Gowtham CP thanks. You mentioned the provided link - which one is that? Not following whether I should still try using Azure Web Apps or App Services to do this.
Gowtham CP 3,730 Reputation points

2024-06-04T12:44:53.01+00:00

Hey matsuo_basho , I was referring to the link you provided. I found the following tutorial to be a good starting point for Azure Functions: https://learn.microsoft.com/en-us/azure/azure-functions/functions-machine-learning-tensorflow?tabs=bash . You probably don't need Azure Web Apps or App Services right now. Kindly refer to my previous comment for more details. Thanks!
matsuo_basho 10 Reputation points

2024-06-14T00:13:59.3466667+00:00

@Gowtham CP Sorry, I'm a bit confused. In a previous post you recommended not using Azure Functions - "skip Azure Functions and ML for now".I've also taken a look at the Azure Functions tutorial, and I'm not sure that it's really applicable in my use case. I will reiterate what I'm trying to do - I have a Plotly Dash app that I can run locally on my machine. The app allows a user to load in data, creates an ML model trained on part of the data, and then outputs a graph of the prediction on the test set.

I want to be able to run this same Dash App on my own website.

From your prior recommendation:
"deploy using a web server or hosting service" it appears that Microsoft / Azure is not where I should turn to?

The problem with going to a web server / hosting service is that the website is actually run on Duda. They told me that I can add the dashboard to my site as long as I can embed it using HTML/JS codes.

Thank you.
Gowtham CP 3,730 Reputation points

2024-06-18T09:19:59.5166667+00:00

Hi matsuo_basho ,

No worries about the Azure ML confusion. It's not the best fit for your project.

Use the dash-uploader component in your Dash app for users to upload their data easily.

Skip Azure ML for now and use your preferred ML library (like scikit-learn or TensorFlow) directly in your Dash app's backend for training and prediction. Once data is uploaded, the backend will handle training the model, making predictions, and preparing the results for visualization.

For deployment on Duda, choose a web server or hosting service like Gunicorn or Heroku, capable of deploying Python applications.

By utilizing dash-uploader, handling training and predictions in your app, and deploying on a Duda-compatible service, you can keep things simple and effective. Thank you!

1 answer

Gowtham CP 3,730 Reputation points

2024-05-31T14:39:31.3833333+00:00
Hello matsuo_basho ,

Thank you for reaching out on the Microsoft Q&A.

1. Set Up Your Website:

Install libraries: Run pip install dash azureml-sdk to get the Dash and Azure ML libraries.

Deploy your app: Use Git or Docker to deploy your app.

2. Connect to Azure ML:

Create a workspace: Set up an Azure ML workspace and train your model.

Link your app to the model: Use the Azure ML SDK to connect your app to the trained model and display predictions.

Helpful Links:

Deploying Dash on Azure

Building a Dashboard with Dash and Azure

I hope this helps! If you have any further questions, feel free to ask.

If the information is useful, please accept the answer and upvote it to assist other community members.
Please sign in to rate this answer.
matsuo_basho 10 Reputation points

2024-05-31T17:35:09.8566667+00:00

@Gowtham CP thanks for the response. The first article is very useful. Will that setup using Azure Web App Services allow the user to load a file onto the dashboard, though? That seems like a little bit more of a complex step that may require some other tools.

Gowtham CP 3,730 Reputation points

2024-05-31T17:46:56.8666667+00:00

Hello matsuo_basho ,

Enabling file upload in your Azure ML dashboard requires additional tools. You'll need a file upload component in your Dash app, Azure Functions for backend processing (storing files, triggering ML pipelines, and retrieving predictions), and integration with Azure ML using the SDK.

Thanks!!

matsuo_basho 10 Reputation points

2024-06-02T15:53:50.71+00:00

error duplicate comment
Sign in to comment

Share via

Hosting a dashboad on my own website

1 answer